The development of exoskeletons control algorithms is a challenging task due to the inevitable interaction between robot and human. Data-driven approaches represent a promising alternative and complement to traditional methods, offering new avenues for discerning user intentions and enhancing controller design. Machine learning models have shown significant efficacy in interpreting data from wearable sensors, leading to advancements in this field. In the present work, we present a Convolutional and Recurrent Neural Network (CRNN) to predict continuous gait phase (GP) based on data coming from a single shank-mounted intertial measurment unit (IMU). Eleven healthy subject were enrolled in the training data collection. The validation employed a verified IMU-based algorithm that detects four primary gait events: heel strike, mid-stance, toe-off, and midswing. Fourteen healthy subjects participated in the validation. The CRNN accurately predicted the continuous GP with an overall RMSE of 0.7% throughout the gait cycle. Compared to the validation algorithm, the CRNN achieved a mean difference of 0.4% and a standard deviation of 3.2% with the algorithm output within all four gait phases, best performing on mid-swing. The results demonstrate the feasibility of the proposed method for implementation in an exoskeleton control algorithm.
A Convolutional and Recurrent Neural Network-Based Control Algorithm for ankle exoskeleton: Validation of performance using IMU-based gait analysis / Liguori, Lorenzo; D'Alvia, Livio; Del Prete, Zaccaria; Palermo, Eduardo. - (2024). (Intervento presentato al convegno 2024 IEEE International Workshop on Metrology for Industry 4.0 & IoT (MetroInd4.0 & IoT) tenutosi a Firenze) [10.1109/metroind4.0iot61288.2024.10584124].
A Convolutional and Recurrent Neural Network-Based Control Algorithm for ankle exoskeleton: Validation of performance using IMU-based gait analysis
Liguori, Lorenzo;D'Alvia, Livio;Del Prete, Zaccaria;Palermo, Eduardo
2024
Abstract
The development of exoskeletons control algorithms is a challenging task due to the inevitable interaction between robot and human. Data-driven approaches represent a promising alternative and complement to traditional methods, offering new avenues for discerning user intentions and enhancing controller design. Machine learning models have shown significant efficacy in interpreting data from wearable sensors, leading to advancements in this field. In the present work, we present a Convolutional and Recurrent Neural Network (CRNN) to predict continuous gait phase (GP) based on data coming from a single shank-mounted intertial measurment unit (IMU). Eleven healthy subject were enrolled in the training data collection. The validation employed a verified IMU-based algorithm that detects four primary gait events: heel strike, mid-stance, toe-off, and midswing. Fourteen healthy subjects participated in the validation. The CRNN accurately predicted the continuous GP with an overall RMSE of 0.7% throughout the gait cycle. Compared to the validation algorithm, the CRNN achieved a mean difference of 0.4% and a standard deviation of 3.2% with the algorithm output within all four gait phases, best performing on mid-swing. The results demonstrate the feasibility of the proposed method for implementation in an exoskeleton control algorithm.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.